Completeness Conditions of a Class of Pattern Recognition Algorithms Based on Image Equivalence

摘要

The paper presents recent results in establishing existence conditions of a class of efficient algorithms for image recognition problem including the algorithm that correctly solves this problem. The proposed method for checking on satisfiability of these conditions is based on the new definition of image equivalence introduced for a special formulation of an image recognition problem. It is shown that the class of efficient algorithms based on estimate calculation contains the correct algorithm in its algebraic closure. The main result is an existence theorem. The obtained theoretical results will be applied to automation of lymphoid tumor diagnostics by the use of hematological specimens.